vsapsai added a comment.
Curiously, the results for the second project are unexpected and bizarre.
| **File** | **Baseline (s)** | **Set Dedupe (s)** | **No external (s)**
| **Set Dedupe (percentage of baseline)** | **No external (percentage of
baseline)** |
| Project2. File1 | 0.378582125 | 0.318364375 |
0.318319875 | 84.094% | 84.082%
|
| Project2. File2 | 0.245813125 | 0.24931675 |
0.25345175 | 101.425% | 103.107%
|
| Project2. File3 | 0.24643325 | 0.260554875 |
0.261508 | 105.730% | 106.117%
|
| Project2. File4 | 0.24745525 | 0.256896 | 0.253167625
| 103.815% | 102.308%
|
| Project2. File5 | 0.2483135 | 0.256163875 |
0.2635545 | 103.161% | 106.138%
|
|
Standard deviation goes down again and doesn't exceed 0.02. This project is a
big desktop app, so expect it to have long-isa dependency chains. Also,
anecdotally, when I compared baseline and "no external" full project builds,
"no external" was 20% faster.
My understanding is that due to the short compilation times the measurements
are less reliable (haven't done anything heroic to make sure nothing else was
running on my machine). And maybe these files are inherently more volatile
because when I was selecting them for measurements, their compile time was >20s
and now we have ~0.25s (with clean module cache it is still below 20s).
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D109632/new/
https://reviews.llvm.org/D109632
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
